Reference

It is all about Fit class of unsatfit.

Methods

For optimization.

set_model(model, const=[])

Set hydraulic model for model with constant parameters of const.

optimize()

Optimize parameteres.

f_ht(p, x)

Water retention function with free parameters p and pressure heads x. To be accessed after set_model.

f_hk(p, x)

Hydraulic conductivity with free parameters p and pressure heads x. To be accessed after set_model.

get_init()

Get initial estimate of water retention parameters except qs and qr. To be accessed after set_model. See models for function names by specifying a model.

get_wrf()

Get water retention parameters. To be accessed after set_model. See models for function names by specifying a model.

modified_model(hs)

Change the hydraulic model to modified model with hs value by Vogel et al. (2000).

For figures etc.

plot()

Plot a figure.

add_curve()

Add a curve for plot.

clear_curves()

Clear curves.

contour(x,y)

Draw contour of RMSE for x and y in parameter name.

test()

Test integrity of the code for development.

Properties for settings

swrc

(h, θ) dataset; h and θ as list respectively

unsat

(h, K) dataset (optimize only WRF when empty list [] is provided)

ini

Initial parameters. When it is a list (or tuple) of parameters, such as [a, b, c], the set of parameters is used as initial parameters. When it is a nested list, such as [[a1, a2, a3], [b1, b2]], all combination of the initial parameters is used.

• For boundary conditions, see source.
• For least square optimization, see source.
• For figure options, see source.

For specifying font in figures, FontProperties object can be set as self.fp. For example, set font_path as the absolute path to a font file and

import unsatfit
from matplotlib.font_manager import FontProperties
f = unsatfit.Fit()
f.fp = FontProperties(fname=font_path, size=9)

It can be used for using Japanese font.

Properties for return values of optimize()

fitted

Fitted parameters

message

Return message

success

True if optimization succeeded

mse_ht, mse_ln_hk

Mean squared error for θ and ln(K), respectively

se_ht, se_ln_hk

Standard error σ for θ and ln(K), respectively

r2_ht, r2_ln_hk

Coefficient of determination (R²) for θ and ln(K), respectively

aic_ht, aic_ln_hk

AIC = 2n ln(σ) + 2k for θ and ln(K), respectively

aicc_ht, aicc_ln_hk

Corrected AIC = AIC + 2k(k+1)/(n-k-1) for θ and ln(K), respectively

jac

Modified Jacobian matrix. For WRF only.

perr

1 σ uncertainty on fitted parameters. For WRF only.

cor

Correlation matrix. For WRF only.